Mining subsidence data space-time fusion method and system based on multiple modes

Abstract

The invention relates to the field of mine data monitoring, in particular to a space-time fusion method and system based on multi-modal mining subsidence data, and the method comprises the steps: collecting mine subsidence information through a sensor cluster, obtaining a collected data set, carrying out the signal preprocessing of the collected data set, and obtaining a time calibration signal set; the time calibration signal set is uploaded to an edge processing device to obtain a temporary storage data set, a field data set in the temporary storage data set is identified, the field data set is transmitted to field equipment to obtain a field data set, the temporary storage data set is uploaded to a centralized processing center to obtain a to-be-processed data set, and the temporary storage data set is deleted based on the edge processing device to obtain the to-be-processed data set. And carrying out space-time fusion on the to-be-processed data set to obtain an initial fusion data set. According to the method, more accurate monitoring of mining subsidence is realized by fusing the spatial-temporal characteristics of the multi-dimensional data.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of mine data monitoring, in particular to a multi-modal mining subsidence data spatio-temporal fusion method and system. BACKGROUND

[0002] Mine subsidence disaster monitoring is the core link of mine safety. Usually, sensors are used for detection, and multiple sensors are set at multiple monitoring points to assist in monitoring.

[0003] The traditional method relies on a single sensor, although it can realize the monitoring of the mining area, there are problems such as time and space coverage limitations, multi-source data asynchrony, etc. The data obtained by single-dimensional monitoring is not comprehensive enough for the evaluation of the mining area, and the multi-dimensional detection data is not synchronized in time and space, and the data volume is large. The data analysis of the monitoring of the mining area brings trouble.

[0004] Therefore, a method is needed that can establish a high-precision time reference according to the characteristics of the sensor, use time as the link between each sensor, align the data of each sensor in time, and analyze the mining subsidence data, so as to fuse the spatio-temporal characteristics of multi-dimensional data and realize more accurate monitoring of mining subsidence. SUMMARY

[0005] The present application provides a multi-modal mining subsidence data spatio-temporal fusion method and computer readable storage medium, which mainly aims to fuse the spatio-temporal characteristics of multi-dimensional data and realize more accurate monitoring of mining subsidence.

[0006] To achieve the above purpose, the present application provides a multi-modal mining subsidence data spatio-temporal fusion method, which comprises: Constructing a sensor network and extracting a sensor cluster from the sensor network, wherein the sensor cluster comprises a plurality of sensors; Collecting mine subsidence information using the sensor cluster and preprocessing the mine subsidence information to obtain a digital signal set, wherein the digital signal set comprises a discrete signal set and a continuous signal set; Performing timestamp calibration on the digital signal set to obtain a time-calibrated signal set; Uploading the time-calibrated signal set to a pre-constructed edge processing device to obtain a temporary data set, identifying the on-site data set in the temporary data set, transmitting the on-site data set to a pre-constructed on-site device to obtain the on-site data set, and uploading the temporary data set to a pre-constructed centralized processing center to obtain a to-be-processed data set; Deleting the temporary data set based on the edge processing device to obtain a cleaned edge processing device; Performing spatio-temporal fusion on the to-be-processed data set to obtain an initial fusion data set; The pre-constructed abnormal data identification model is used for data cleaning on the initial fusion data set to obtain an optimized fusion data set, and the pre-constructed visualization model set is updated according to the optimized fusion data set; Based on the updated visualization model set, the edge cleaning device and the field data set, the spatiotemporal fusion of multi-modal mining subsidence data is completed.

[0007] Optionally, the sensor network is constructed, including: A target monitoring parameter set is obtained, and a region space model corresponding to the target monitoring parameter set is confirmed. Target monitoring parameters are sequentially extracted from the target monitoring parameter set, and the following operations are performed on the extracted target monitoring parameters: One or more key layout points are confirmed from the region space model, a plurality of placeable points are identified from the region space model based on a preset control distance control parameter, key layout points are sequentially extracted from the one or more key layout points, and the following operations are performed on the key layout points: The key layout points are combined with the plurality of placeable points to obtain a plurality of matching groups, and a spatial Euclidean distance of each matching group in the plurality of matching groups is calculated to obtain a distance set; The distance set is summarized, and all distances greater than the control distance control parameter are extracted from the summarized distance set to obtain a plurality of suboptimal distances. Suboptimal control points corresponding to each suboptimal distance in the plurality of suboptimal distances are identified to obtain a plurality of suboptimal control points; Based on a preset sensor layout density gradient, one or more key layout points and a plurality of suboptimal control points are used to construct a plurality of precision layout groups. The plurality of precision layout groups are matched according to a preset time period precision requirement to obtain a plurality of precision requirement matching groups. A single sensor network based on dynamically adjusted collection precision based on a time period is constructed according to the plurality of precision requirement matching groups, and the single sensor network is summarized to obtain a sensor network corresponding to the target monitoring parameter set.

[0008] Optionally, based on the preset sensor layout density gradient, one or more key layout points and a plurality of suboptimal control points are used to construct a plurality of precision layout groups, including: A target precision gradient is extracted from a sensor layout density gradient, wherein the sensor layout density gradient includes a high-density gradient, a medium-density gradient, and a low-density gradient; Based on the target precision gradient, a plurality of target sensors for monitoring the target monitoring parameter are confirmed, and a plurality of key sensors for being arranged at the one or more key layout points are extracted from the plurality of target sensors to obtain a key sensor group, wherein the key sensor group includes one or more key sensors, and the key sensors correspond one-to-one to the key layout points; The key sensor group is removed from the plurality of target sensors to obtain a plurality of remaining sensors. A total number of traversal strategies is calculated based on the plurality of remaining sensors and the plurality of suboptimal control points. A plurality of traversal layout strategies is determined based on the total number of traversal strategies, wherein the total number of traversal strategies is equal to a total number of the plurality of traversal layout strategies, and the total number of traversal strategies is calculated according to the following formula:

[0009] wherein, the total number of traversal strategies, the total number of the plurality of suboptimal control points, the total number of the plurality of remaining sensors, denotes a factorial symbol; The traversal layout strategy is extracted from the plurality of traversal layout strategies in sequence. A detection range coverage, a sensor accuracy average, and a network delay average of the traversal layout strategy are calculated to obtain a strategy evaluation parameter. A strategy evaluation value of the traversal layout strategy is calculated based on a pre-constructed evaluation method and the strategy evaluation parameter. The strategy evaluation values are summarized, and one or more strategy evaluation values with the largest values are extracted from the summarized strategy evaluation values to obtain one or more suboptimal evaluation values. If there are a plurality of suboptimal evaluation values with the same values, a plurality of dispersions corresponding to the plurality of suboptimal evaluation values with the same values are calculated. The traversal layout strategy corresponding to the suboptimal evaluation value with the largest dispersion in the plurality of dispersions is determined as the suboptimal control point layout strategy of the target accuracy gradient. If there is one suboptimal evaluation value, the traversal layout strategy corresponding to the suboptimal evaluation value is determined as the suboptimal control point layout strategy of the target accuracy gradient. The accuracy layout group of the target accuracy gradient is generated based on the suboptimal control point layout strategy and the one or more key layout points. The accuracy layout groups are summarized to obtain a plurality of accuracy layout groups.

[0010] Optionally, the sensor cluster is used to collect mine subsidence information, and the mine subsidence information is preprocessed to obtain a digital signal set, including: The target sensor is extracted from the sensor cluster in sequence, and the following operations are performed on the extracted target sensor: The pre-constructed target sampling information is signal collected based on a preset sensor sampling frequency of the target sensor, to obtain a collection signal, and a pre-constructed wavelet transform filtering operation is used to denoise the collection signal to obtain a denoising signal, wherein the wavelet transform filtering operation includes a plurality of modified filtering calibration parameters, and the plurality of modified filtering calibration parameters are obtained by using a pre-constructed orthogonal experiment method to perform a parameter modification operation on a plurality of pre-constructed filtering calibration parameters based on the target sensor, and the plurality of modified filtering calibration parameters are respectively: a wavelet basis function, a wavelet threshold, and a decomposition layer number. The denoising signals are summarized to obtain a denoising signal set corresponding to the sensor cluster, and a signal classification operation is performed on the denoising signal set based on signal continuity discrimination to obtain a digital signal set.

[0011] Optionally, the digital signal set is executed with a timestamp calibration to obtain a time calibration signal set, including: A discrete signal set is identified from the digital signal set, and the discrete signal set is divided based on the sensor cluster to obtain a discontinuous numerical group set, and each discontinuous numerical value in the discontinuous numerical group set is executed with a timestamp identification to obtain an identified discontinuous numerical group set, wherein the timestamp identification includes sensor identification, monitoring point identification, and time identification; A continuous digital signal set corresponding to the continuous signal set is obtained, and the continuous digital signal set is executed with a timestamp calibration based on the identified discontinuous numerical group set to obtain an identified continuous numerical group set; The identified continuous numerical group set and the identified discontinuous numerical group set are summarized to obtain a time calibration signal set.

[0012] Optionally, the continuous digital signal set is executed with a timestamp calibration based on the identified discontinuous numerical group set to obtain an identified continuous numerical group set, including: A time identification set is extracted from the identified discontinuous numerical group set, a time start point and a time end point in the time identification set are identified, a continuous time axis is constructed according to the time start point and the time end point, and an anchor point is performed on the continuous time axis by using the time identification in the time identification set to obtain an identified time axis, wherein the identified time axis includes a plurality of time points. Based on the plurality of time points, each continuous digital signal in the continuous digital signal set is discretized based on the time points to obtain a discrete signal group set, wherein the discrete signal group corresponds to the continuous digital signal one by one, and the discrete signal group includes a plurality of continuous-discrete signals. The discrete signal group set is executed with a timestamp calibration to obtain an identified continuous numerical group set.

[0013] Optionally, the temporary data set is deleted by the edge processing device to obtain a cleaned edge processing device, including: judging whether the edge processing device completes a preset transmission task, wherein the transmission task is an operation of transmitting the live data set to the pre-built live device, obtaining the live data set, uploading the temporary data set to the pre-built centralized processing center, and obtaining the to-be-processed data set; if it is confirmed that the edge processing device completes the transmission task, performing data integrity verification on the live data set and the to-be-processed data set to obtain an integrity verification result; if the integrity verification result is a preset correct transmission, monitoring a total storage time length of the temporary data set in the edge processing device, and if the total storage time length reaches a preset temporary storage time length, deleting the temporary data set in the edge processing device to obtain a cleaned edge processing device.

[0014] Optionally, the cleaning of the edge processing device based on the deletion of the temporary data set further includes: if it is confirmed that the edge processing device does not complete the transmission task, returning to the operation of identifying the live data set in the temporary data set, and performing an increment operation on a pre-built transmission counter to obtain a transmission number, wherein an initial value of the transmission counter is 0, and the increment operation on the transmission counter is performed once the operation of returning to the identification of the live data set in the temporary data set is performed; until it is confirmed that the edge processing device completes the transmission task, the transmission number is greater than or equal to a preset repetition threshold, or the integrity verification result is a preset incorrect transmission; when the integrity verification result is the preset incorrect transmission, returning to the operation of returning to the identification of the live data set in the temporary data set and performing the increment operation on the pre-built transmission counter; when the transmission number is greater than or equal to the preset repetition threshold, generating a communication error warning instruction.

[0015] Optionally, the data cleaning on the initial fusion data set based on the pre-built abnormal data identification model to obtain an optimized fusion data set includes: extracting initial fusion data from the initial fusion data set in sequence, and performing the following operations on the extracted initial fusion data: performing data abnormal value identification on the initial fusion data to obtain an abnormal data group, wherein the abnormal data group includes zero, one or more abnormal data; if the abnormal data group is not an empty set, performing abnormal value elimination on the initial fusion data based on the abnormal data group to obtain intermediate fusion data containing one or more missing values; The intermediate fusion data and the pre-constructed mask matrix are taken as input parameters and imported into a pre-constructed missing value generator, the output parameters corresponding to the input parameters are predicted by using the missing value generator, and suboptimal fusion data containing one or more predicted interpolation values are obtained; The suboptimal fusion data is imported into a pre-constructed discriminator, the interpolation correct probability of each predicted interpolation value in the suboptimal fusion data is estimated by using the discriminator, one or more interpolation correct probabilities are obtained, the predicted interpolation values corresponding to all interpolation correct probabilities greater than a preset probability threshold in the one or more interpolation correct probabilities are taken as available interpolation values, the intermediate fusion data is interpolated, and optimal intermediate fusion data is obtained; The mask matrix is adjusted based on the optimal intermediate fusion data, an updated adjusted mask matrix is obtained, the optimal intermediate fusion data is taken as the intermediate fusion data, the updated adjusted mask matrix is taken as the mask matrix, and the step of taking the intermediate fusion data and the pre-constructed mask matrix as input parameters and importing them into the pre-constructed missing value generator is returned until the optimal intermediate fusion data does not contain missing values, and optimal fusion data is obtained; The optimal fusion data is summarized to obtain an optimal fusion data set.

[0016] To achieve the above object, the application further provides a multi-modal mining subsidence data spatio-temporal fusion system, comprising: A preprocessing data module is configured to construct a sensor network and extract a sensor cluster from the sensor network, wherein the sensor cluster comprises a plurality of sensors, the sensor cluster is used to collect mine subsidence information, and the mine subsidence information is preprocessed to obtain a digital signal set, wherein the digital signal set comprises a discrete signal set and a continuous signal set; A time calibration module is configured to perform timestamp calibration on the digital signal set to obtain a time-calibrated signal set; An uploading module is configured to upload the time-calibrated signal set to a pre-constructed edge processing device to obtain a temporary data set, identify an on-site data set in the temporary data set, transmit the on-site data set to a pre-constructed on-site device to obtain an on-site data set, and upload the temporary data set to a pre-constructed centralized processing center to obtain a to-be-processed data set, delete the temporary data set based on the edge processing device, and obtain a cleaned edge processing device; A post-processing module is configured to perform spatio-temporal fusion on the to-be-processed data set to obtain an initial fusion data set, perform data cleaning on the initial fusion data set by using a pre-constructed abnormal data identification model to obtain an optimal fusion data set, update a pre-constructed visualization model set based on the optimal fusion data set, and complete the multi-modal mining subsidence data spatio-temporal fusion based on the updated visualization model set, the cleaned edge processing device, and the on-site data set.

[0017] To solve the above problems, the application further provides an electronic device, which comprises: a memory storing at least one instruction; and a processor executing the instruction stored in the memory to realize the multi-modal mining subsidence data spatio-temporal fusion method.

[0018] To solve the above problems, the application further provides a computer readable storage medium, which stores at least one instruction, and the at least one instruction is executed by a processor in an electronic device to realize the multi-modal mining subsidence data spatio-temporal fusion method.

[0019] To solve the problems described in the background art, the present application uses a sensor cluster to collect data of different modalities, jointly describes the state of the subsidence of the mining area, and constructs a sensor network to more comprehensively monitor the mining area corresponding to the regional space, and uses a single sensor network that dynamically adjusts the collection precision under the condition of meeting the precision requirement, and under the condition of low precision requirement, some sensors are turned off, thereby saving energy consumption. The digital signal set is converted into a digital signal set through digital-to-analog conversion, and the digital signal set is time-stamped to obtain a time-stamped signal set. It can be seen that the present application processes digital signals, because digital signals are helpful for anti-interference, high precision and programmable processing through binary coding, and provide a standardized data basis for subsequent storage, transmission and intelligent analysis. The digital signal set is time-stamped to obtain a time-stamped signal set, and the present application represents different sensors in a mining area at the same time point and the time development law through the identification time axis. The time-stamped signal set is uploaded to the pre-constructed edge processing device to obtain a temporary storage data set, the on-site data set in the temporary storage data set is identified, the on-site data set is transmitted to the pre-constructed on-site device to obtain the on-site data set, and the temporary storage data set is uploaded to the pre-constructed centralized processing center to obtain a to-be-processed data set. It can be seen that the present application can temporarily store data and increase the stability of data transmission. The edge processing device deletes the temporary storage data set to obtain a cleaned edge processing device. The present application also performs data integrity verification to check whether the data conforms to the pre-defined format and ensure that the time series data has no breakpoints or out-of-order. The present application performs an add-one operation on the transmission count to obtain the transmission frequency, whether the integrity verification result is a preset error transmission or the edge processing device does not complete the transmission task. When the transmission frequency is greater than or equal to a preset repetition operation threshold, the present application generates a communication error warning instruction to notify the monitoring personnel that a data transmission error has occurred. Because different sensors are turned on at different time periods, there may be abnormal data in the initial fusion data due to transmission errors of different sensors and on-off fluctuations of the sensors during the processing process. Therefore, the present application also performs data cleaning on the initial fusion data set. The present application realizes more accurate monitoring of mining subsidence by fusing the space-time characteristics of multi-dimensional data. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A flowchart of a multi-modal mining subsidence data space-time fusion method according to an embodiment of the present application is provided. Figure 2 A functional module diagram of a multi-modal mining subsidence data space-time fusion system according to an embodiment of the present application is provided. Figure 3 A structural diagram of an electronic device for implementing the multi-modal mining subsidence data space-time fusion method according to an embodiment of the present application is provided.

[0021] BRIEF DESCRIPTION OF DRAWINGS 1. Electronic device; 10. Processor; 11. Memory; 12. Bus.

[0022] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments in conjunction with the drawings. DETAILED DESCRIPTION

[0023] It should be understood that the specific embodiments described herein are merely exemplary and are not intended to limit the present application.

[0024] An embodiment of the present application provides a multi-modal mining subsidence data spatio-temporal fusion method. An execution subject of the multi-modal mining subsidence data spatio-temporal fusion method includes but is not limited to at least one of electronic devices such as a server and a terminal, which can be configured to execute the method provided by the embodiment of the present application. In other words, the multi-modal mining subsidence data spatio-temporal fusion method can be executed by software or hardware installed in a terminal device or a server device, and the software can be a blockchain platform. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0025] Reference Figure 1 As shown in the figure, the multi-modal mining subsidence data spatio-temporal fusion method provided by an embodiment of the present application is a flowchart. In this embodiment, the multi-modal mining subsidence data spatio-temporal fusion method includes: S1, constructing a sensor network and extracting a sensor cluster from the sensor network, wherein the sensor cluster includes a plurality of sensors.

[0026] It can be understood that the sensor network is a network composed of a sensor cluster, and when a sensor in the sensor cluster is enabled, the information collected by the enabled sensor can be obtained through the sensor network.

[0027] It should be noted that the sensors in the sensor cluster are selected according to the parameters to be monitored. The sensors referred to in the present application include but are not limited to water level sensors, displacement sensors, weather sensors, Beidou sensors, etc. Among them, the water level sensor, the displacement sensor and the weather sensor are respectively used to monitor the corresponding water level, displacement and weather. The Beidou sensor is used to monitor the water vapor inversion.

[0028] Further, the construction of the sensor network includes: Obtaining a target monitoring parameter set and confirming a region space model corresponding to the target monitoring parameter set, sequentially extracting a target monitoring parameter from the target monitoring parameter set, and performing the following operations on the extracted target monitoring parameter: Confirm one or more key layout points from the area space model, identify a plurality of placeable points from the area space model based on a preset control distance control parameter, sequentially extract key layout points from one or more key layout points, and perform the following operations on the key layout points: Combine the key layout points with the plurality of placeable points to obtain a plurality of matching groups, calculate the spatial Euclidean distance of each matching group in the plurality of matching groups to obtain a distance set; Summarize the distance set and extract all distances greater than the control distance control parameter from the summarized distance set to obtain a plurality of suboptimal distances, identify a suboptimal control point corresponding to each suboptimal distance in the plurality of suboptimal distances to obtain a plurality of suboptimal control points; Based on a preset sensor layout density gradient, one or more key layout points and a plurality of suboptimal control points are used to construct a plurality of precision layout groups, and the plurality of precision layout groups are matched according to a preset time period accuracy requirement to obtain a plurality of precision requirement matching groups; According to the plurality of precision requirement matching groups, a single sensor network based on dynamic adjustment of collection accuracy based on time periods is constructed, and the single sensor network is summarized to obtain a sensor network corresponding to the target monitoring parameter set.

[0029] It should be noted that the target monitoring parameter set is a set of a plurality of target monitoring parameters, and the target parameter is a parameter monitored by the sensor in the sensor network, for example, the water level in the A pit, the temperature at the A pit, etc.

[0030] It can be understood that the area space model is a simplified three-dimensional model corresponding to a mining area that needs to be monitored by a sensor network, which is used to reproduce the topography and facility layout of the mining area, and provides a reference for setting points for sensors in the sensor network. Among them, the mining area specifically refers to the local area that needs to be monitored in the entire mining area, for example, if the target monitoring parameter is the ventilation speed of the ventilation machine in the mining area, the mining area should be the area where the ventilation machine is located, and if the target monitoring parameter is the meteorological data of a mine, the mining area should be set to the open part above the actual mining mine. Therefore, for different sensors and different target monitoring parameters, the mining area is different, and the present application will not be exemplified one by one.

[0031] Further, one or more key layout points are the points where sensors should be set when monitoring the target monitoring parameter through expert experience evaluation, for example, if the target monitoring parameter is the ventilation speed of the ventilation machine in the mining area, according to the expert experience evaluation, at least a sensor should be set at the outlet of the ventilation machine, at this time, the outlet of the ventilation machine is the key layout point.

[0032] Specifically, the control distance control parameter is a parameter for controlling the distance between sensors in the region space model. The placeable point is a site where a sensor can be placed. The sensors of one or more key layout points are not affected by the control distance control parameter. The present application divides the region where the sensor can be placed in the region space model by using the control distance control parameter, thereby obtaining a plurality of placeable points. The uniform division can divide the region where the sensor can be placed in the region space model into a grid with the control distance parameter as the length of the side of an equilateral triangle. The placeable points can also be uniformly set in the region where the sensor can be placed in the region space model. Therefore, there are various existing technologies that can be implemented, and the present application does not limit them.

[0033] Further, the step of combining the key layout points with the plurality of placeable points to obtain a plurality of paired groups is: combining the key layout points with the plurality of placeable points two by two, so that there is one key layout point and one placeable point in the paired group. For example, the plurality of placeable points include a first placeable point and a second placeable point, and the plurality of paired groups are respectively paired group

key layout point, first placeable point

key layout point, second placeable point

[0034] Specifically, the spatial Euclidean distance is the straight-line distance between the placeable point and the key layout point in the Euclidean space. The distance set is the operation of collecting the distance sets corresponding to all key layout points.

[0035] It should be noted that the purpose of setting the placeable point is to improve the monitoring accuracy, expand the monitoring range, and enable the sensor network to more comprehensively monitor the mining area corresponding to the region space. When setting the plurality of placeable points, the present application does not consider whether one or more key layout points coincide with the placeable point, and at the same time, the setting of one or more key layout points is not controlled by the control distance control parameter. Therefore, the key layout point and the placeable point may coincide, or the distance between the key layout point and the placeable point may be too small. If such placeable points are used and sensors are set at such placeable points, the original intention of setting the placeable points of the present application cannot be achieved. Therefore, the present application only extracts the placeable points corresponding to the plurality of suboptimal distances as suboptimal control points.

[0036] Specifically, the suboptimal distance is a distance greater than the control distance control parameter, and the suboptimal control point is the placeable point in the paired group corresponding to the suboptimal distance.

[0037] Further, the time period accuracy requirement is different for different time periods, and the accuracy requirement for monitoring the target monitoring parameter. For example, the time period accuracy requirement for monitoring the target monitoring parameter is high accuracy requirement when mining in the mining area, the time period accuracy requirement for monitoring the target monitoring parameter is medium accuracy requirement in 6:00-18:00, and the time period accuracy requirement for monitoring the target monitoring parameter is low accuracy requirement in 18:00-6:00 the next day.

[0038] It should be noted that the accuracy layout group is the layout mode of the sensor in the mining area under the condition of the corresponding sensor layout density gradient. The meaning of the accuracy requirement matching is that the accuracy layout group used for the corresponding accuracy requirement is confirmed according to different accuracy requirements (high accuracy requirement, medium accuracy requirement, and low accuracy requirement mentioned in the above example), and therefore, the accuracy requirement matching group corresponding to the high accuracy requirement can be exemplified as

high accuracy requirement-high density gradient

[0039] It should be noted that the sensor network in the present application includes a plurality of sensors, and if all the sensors are monitored at all times, a large amount of energy consumption will be required, and therefore, the present application expects to save energy consumption by closing some sensors in the case of low accuracy requirement while meeting the accuracy requirement.

[0040] Further, the single-sensor network is a unit sensor network constructed by a plurality of accuracy requirement matching groups, and the single-sensor network controls the start and stop of the sensor in the single-sensor network according to the plurality of accuracy requirement matching groups and the time period accuracy requirement. For example, when monitoring the target monitoring parameter, the high accuracy requirement corresponds to the accuracy layout group using the high density gradient, the medium accuracy requirement corresponds to the accuracy layout group using the medium density gradient, and so on. The unit sensor network for switching the accuracy layout group according to the accuracy requirement of different time periods is the single-sensor network for dynamically adjusting the collection accuracy.

[0041] Further, the aggregation of the single-sensor network to obtain the sensor network corresponding to the target monitoring parameter set means aggregating the single-sensor networks of different target monitoring parameters, and recording the aggregated plurality of single-sensor networks as the sensor network.

[0042] Further, the plurality of accuracy layout groups are constructed based on the preset sensor layout density gradient, using one or more key layout points and a plurality of suboptimal layout points, and the plurality of accuracy layout groups include: extracting a target accuracy gradient from the sensor layout density gradient, wherein the sensor layout density gradient includes: a high density gradient, a medium density gradient, and a low density gradient; Based on the target precision gradient, a plurality of target sensors for monitoring target monitoring parameters are confirmed, a plurality of key sensors for being arranged at the one or more key layout points are extracted from the plurality of target sensors, and a key sensor group is obtained, wherein the key sensor group includes one or more key sensors, and the key sensors correspond to the key layout points one by one; The key sensor group is removed from the plurality of target sensors to obtain a plurality of remaining sensors, a total number of traversal strategies is calculated based on the plurality of remaining sensors and the plurality of suboptimal control points, and a plurality of traversal layout strategies are confirmed based on the total number of traversal strategies, wherein the total number of traversal strategies is equal to the total number of the plurality of traversal layout strategies, and the calculation formula of the total number of traversal strategies is as follows:

[0043] Wherein, The total number of traversal strategies is represented by T, The total number of the plurality of suboptimal control points is represented by N, The total number of the plurality of remaining sensors is represented by M, The factorial symbol is represented by n!; and The traversal layout strategies are sequentially extracted from the plurality of traversal layout strategies, the detection range coverage, the sensor precision average, and the network delay average of the traversal layout strategies are calculated, the strategy evaluation parameters are obtained, the strategy evaluation values of the traversal layout strategies are calculated based on the pre-constructed evaluation method and the strategy evaluation parameters; The strategy evaluation values are summarized, and one or more optimal evaluation values are obtained by extracting the one or more strategy evaluation values with the largest value from the summarized strategy evaluation values, if there are a plurality of optimal evaluation values with the same value, a plurality of dispersions corresponding to the plurality of optimal evaluation values with the same value are calculated, and the traversal layout strategy corresponding to the optimal evaluation value with the largest dispersion in the plurality of dispersions is confirmed as the suboptimal control point layout strategy of the target precision gradient; If there is one optimal evaluation value, the traversal layout strategy corresponding to the optimal evaluation value is confirmed as the suboptimal control point layout strategy of the target precision gradient; The precision layout group of the target precision gradient is generated based on the suboptimal control point layout strategy and the one or more key layout points; The precision layout groups are summarized to obtain a plurality of precision layout groups.

[0044] It can be understood that the high-density gradient, the medium-density gradient, and the low-density gradient are artificial classifications of the density gradient, wherein the number of the plurality of target sensors in the high-density gradient is greater than the number of the plurality of target sensors in the medium-density gradient, and the number of the plurality of target sensors in the medium-density gradient is greater than the number of the plurality of target sensors in the low-density gradient.

[0045] Further, the target precision gradient is an accuracy gradient extracted from the high-density gradient, the medium-density gradient, and the low-density gradient. The number of the target sensors is determined according to the target precision gradient, and different target precision gradients correspond to different numbers of target sensors, which are set by humans. For example, in the case of the water level as the target monitoring parameter, the number of target sensors corresponding to the high-density gradient is 20, the number of target sensors corresponding to the medium-density gradient is 15, and the number of target sensors corresponding to the low-density gradient is 10. Different target monitoring parameters correspond to different numbers of sensors under the same target precision gradient. For example, in the case of the weather as the target monitoring parameter, the number of target sensors corresponding to the high-density gradient can be 30.

[0046] Specifically, the key sensors are target sensors arranged at the one or more key layout points. The remaining sensors are all target sensors remaining after the key sensors are removed from the target sensors. The total number of traversal strategies is used to describe the total number of all possible schemes of arranging the remaining sensors in the suboptimal control points. In the present application, the total number of traversal strategies can be represented by the number of combinations. The traversal layout strategy is a layout strategy represented by a combination number corresponding to the total number of traversal strategies. For example, the total number of suboptimal control points is 4, and the total number of remaining sensors is 2. Therefore, according to the calculation formula of the total number of traversal strategies, the total number of traversal strategies can be calculated as 6, that is, there are 6 ways of arranging 2 remaining sensors in 4 suboptimal control points. One of the 6 ways of arrangement is one of the traversal layout strategies in the plurality of traversal layout strategies. Therefore, there are 6 traversal layout strategies at this time.

[0047] It can be understood that the detection range coverage rate is calculated as follows: the unit detection range of the remaining sensors is obtained, and all the repeated detection ranges corresponding to the remaining sensors arranged based on the traversal layout strategy are obtained. The overlapping part of each remaining sensor unit detection range and the repeated detection range is subtracted to obtain the non-repeated detection range of each remaining sensor. The area sum of all non-repeated detection ranges and the area of the repeated detection range are calculated. The area sum of the non-repeated detection range and the area of the repeated detection range are added to obtain the detection range of the traversal layout strategy. The detection range of the traversal layout strategy is divided by the area of the mining area to obtain the detection range coverage rate. The repeated detection range is a range where multiple remaining sensors repeatedly detect, and the non-repeated detection range is a range where only one remaining sensor can detect.

[0048] Furthermore, the detection range coverage rate represents the ratio of the detection range of the traversal layout strategy to the area of ​​the mining area. The higher the detection range coverage rate, the wider the mining area that the traversal layout strategy can cover, and thus the higher the monitoring accuracy.

[0049] Furthermore, the formula for calculating the average sensor accuracy is as follows:

[0050] in, This represents the average accuracy of the sensor. Indicates the first The unit test accuracy of the remaining sensors, This indicates the total number of remaining sensors. Indicates the first Remaining sensors Secondary sensor accuracy test Indicates the first The total number of times the remaining sensors underwent sensor accuracy testing. Indicates the first Remaining sensors The true value of the sensor accuracy test. Indicates the first Remaining sensors The test value of the sensor accuracy test performed this time.

[0051] It should be noted that the unit test precision is the first... The test accuracy of the remaining sensors is therefore the average of the unit test accuracy of each of the multiple remaining sensors. This invention uses the average sensor accuracy to characterize the test accuracy of the traversal layout strategy.

[0052] Furthermore, the aforementioned sensor accuracy test refers to: given the true value, using the remaining sensors to detect the test value, and characterizing the error of a single sensor accuracy test by the absolute value of the relative error between the true value and the test value, thus obtaining the single-shot error (i.e., In obtaining the unit test accuracy of a remaining sensor, this invention performs multiple sensor accuracy tests, obtaining multiple single errors. The average of these multiple single errors is taken as the mean error of the remaining sensor (i.e., ...). ), thereby calculating the unit test accuracy of the remaining sensor (i.e. ).

[0053] Further, the network delay average is an average of network delays of each of the remaining sensors, and the network delay of each of the remaining sensors can be obtained by a network delay test, and a calculation formula for calculating the network delay in the network delay test is:

[0054] wherein, represents the network delay, represents a test duration in the network delay test, represents a transmission number of the remaining sensor in the test duration.

[0055] It should be understood that the test duration refers to a time difference between a start time and an end time of the test in the network delay test. The transmission number refers to a number of times that the remaining sensor can send data from the sensor itself to the corresponding single-sensor network in the test duration.

[0056] It can be understood that the present application considers that the lower the network delay average, the more the measurement error can be reduced when monitoring by using the traversal layout strategy, and the higher the quality of the collected data can be improved. Therefore, when the network delay average is lower, the present application is more inclined to select the traversal layout strategy corresponding to the network delay average.

[0057] It can be understood that the present application takes the detection range coverage, the sensor accuracy average and the network delay average of each traversal layout strategy as a strategy evaluation parameter for evaluating the traversal layout strategy, and considers that the larger the detection range coverage, the larger the sensor accuracy average, and the smaller the network delay average, the better the corresponding traversal layout strategy. Therefore, the present application uses a pre-constructed evaluation method to calculate a strategy evaluation value of the strategy evaluation parameter. Optionally, the pre-constructed evaluation method can be an analytic hierarchy process or an entropy weight method, and various existing technologies can be implemented, and thus will not be described here.

[0058] Specifically, the strategy evaluation value is a numerical value for evaluating the advantages and disadvantages of the traversal layout strategy.

[0059] Further, the suboptimal evaluation value is the strategy evaluation value with the largest value in the aggregated strategy evaluation values. If there are multiple strategy evaluation values with the same value and the value is the largest in the aggregated strategy evaluation values, the application can obtain multiple suboptimal evaluation values. When there are multiple suboptimal evaluation values, the multiple dispersion degrees corresponding to the multiple suboptimal evaluation values with the same value are calculated, wherein the dispersion degree and the suboptimal evaluation value with the same value correspond to each other. The dispersion degree is the dispersion degree between the multiple remaining sensor setting points in the traversal layout strategy. The application can be characterized by the sum of the distances between two sensors, and many other prior art can calculate the dispersion degree, which will not be exemplified here. The application takes the dispersion degree into consideration, and expects to maximize the spatial distribution uniformity of the single-sensor network, thereby improving the monitoring coverage of the single-sensor network. Therefore, the traversal layout strategy corresponding to the suboptimal evaluation value with the largest dispersion degree is confirmed as the suboptimal control point layout strategy of the target precision gradient. The suboptimal control point layout strategy is the layout strategy required to be applied by the multiple remaining sensors in the actual mining area.

[0060] Further, the target precision gradient precision layout group based on the suboptimal control point layout strategy and the one or more key layout points refers to the process of combining the suboptimal control point layout strategy and the one or more key layout points to obtain the precision layout group, that is, the precision layout group of the target precision gradient includes the suboptimal control point layout strategy and the one or more key layout points.

[0061] S2, collecting mine subsidence information by using a sensor cluster, and preprocessing the mine subsidence information to obtain a digital signal set, wherein the digital signal set includes a discrete signal set and a continuous signal set.

[0062] It can be understood that the mine subsidence information is a collection of information collected by the sensor network in a past period of time, and the past period of time can be one day, one week, or one month, etc., which is not limited by the application.

[0063] It can be understood that the collection of mine subsidence information by using a sensor cluster and the preprocessing of the mine subsidence information to obtain a digital signal set include: The target sensor is extracted from the sensor cluster in sequence, and the following operations are performed on the extracted target sensor: The pre-constructed target sampling information is signal collected based on a preset sensor sampling frequency of the target sensor, to obtain a collection signal, and the collection signal is denoised by using a pre-constructed wavelet transform filtering operation, to obtain a denoised signal, wherein the wavelet transform filtering operation includes a plurality of modified filtering calibration parameters, and the plurality of modified filtering calibration parameters are obtained by using a pre-constructed orthogonal experiment method to perform a parameter modification operation on a plurality of filtering calibration parameters based on the target sensor, and the plurality of modified filtering calibration parameters are respectively a wavelet base function, a wavelet threshold value, and a decomposition layer number. The denoised signals are summarized to obtain a denoised signal set corresponding to the sensor cluster, and a signal classification operation is performed on the denoised signal set based on signal continuity discrimination to obtain a digital signal set.

[0064] It should be noted that the sensor sampling frequency is the sampling frequency of the target sensor. The target sampling information is the object of the target monitoring parameter required by the target sensor, for example, if the target sensor is a temperature sensor, the target monitoring parameter is temperature, and the target sampling information is the temperature at the air inlet A of the ventilation device in the mining area. The signal collection is a process of collecting signals by using a sensor.

[0065] Specifically, the wavelet transform filtering operation is an operation of filtering based on wavelet transform, which is a prior art and will not be described here. The denoised signal is obtained by using the pre-constructed wavelet transform filtering operation to filter the collection signal to achieve the purpose of denoising.

[0066] It should be noted that there are a plurality of filtering calibration parameters in the wavelet transform, and the plurality of filtering calibration parameters need to be adjusted for the target sensor when the collection signal is denoised by using the wavelet transform. Therefore, the plurality of modified filtering calibration parameters are determined by using the orthogonal experiment method before the collection signal is denoised, and the process of confirming the parameters by using the orthogonal experiment method can be realized by using the prior art, which will not be described here.

[0067] It can be understood that the wavelet base function is a function generated by scaling and translating a mother wavelet, which is used for multi-scale decomposition and time-frequency analysis of an initial digital signal. Different wavelet base functions have different sensitivities to signal characteristics, so the wavelet base function needs to be confirmed according to the target sensor. The wavelet threshold value is a preset threshold value for distinguishing the main signal from the noise in the wavelet transform. The decomposition layer number is the iteration number of the signal decomposition in the wavelet transform process.

[0068] It should be noted that due to the difference in sensor types, the collected denoised signals are not necessarily continuous signals or discrete signals, so the signal classification operation based on signal continuity discrimination is performed to record the continuous denoised signals as continuous signals and the discontinuous signals as discrete signals.

[0069] S3, performing timestamp calibration on the digital signal set to obtain a time-calibrated signal set.

[0070] Further, the performing timestamp calibration on the digital signal set to obtain a time-calibrated signal set comprises: identifying a discrete signal set from the digital signal set, dividing the discrete signal set based on a sensor cluster to obtain a discontinuous numerical group set, and performing timestamp identification on each discontinuous numerical value in the discontinuous numerical group set to obtain an identified discontinuous numerical group set, wherein the timestamp identification comprises sensor identification, monitoring point identification, and time identification; obtaining a continuous digital signal set corresponding to a continuous signal set, and performing timestamp calibration on the continuous digital signal set based on the identified discontinuous numerical group set to obtain an identified continuous numerical group set; summarizing the identified continuous numerical group set and the identified discontinuous numerical group set to obtain the time-calibrated signal set.

[0071] Understandably, the dividing the discrete signal set based on a sensor cluster to obtain a discontinuous numerical group set means that the discontinuous digital signal is divided according to different sensor clusters, so that all discontinuous numerical values in the discontinuous numerical group set after division come from the same sensor.

[0072] Specifically, the performing timestamp identification is an operation of utilizing timestamp identification on discontinuous numerical values, which can be realized by a key-value pair, text identification, or a pointer, and the present application does not limit this. The sensor identification, the monitoring point identification, and the time identification respectively represent an identification composed of a sensor model, target sampling information, and time corresponding to the identified discontinuous numerical value.

[0073] Further, the performing timestamp calibration on the continuous digital signal set based on the identified discontinuous numerical group set to obtain an identified continuous numerical group set comprises: extracting a time identification set from the identified discontinuous numerical group set, identifying a time start point and a time end point in the time identification set, constructing a continuous time axis according to the time start point and the time end point, and anchoring the time identification in the continuous time axis by utilizing the time identification in the time identification set to obtain an identified time axis, wherein the identified time axis comprises a plurality of time anchor points. based on the plurality of time anchor points, discretizing each continuous digital signal in the continuous digital signal set based on the time anchor points to obtain a discrete signal group set, wherein each discrete signal group in the discrete signal group set corresponds to a continuous digital signal, and the discrete signal group comprises a plurality of continuous-discrete signals. performing timestamp calibration on the discrete signal group set to obtain the identified continuous numerical group set.

[0074] Specifically, the time identifier set is a set of identifiers identifying all time identifiers in the discontinuous numerical value group set. The continuous time axis is a time axis with the time start point and the time end point as the start point and the end point of the continuous time axis, and the continuous time axis is a one-dimensional continuous time axis. The use of the time identifier in the time identifier set to anchor the continuous time axis includes identifying the corresponding position of the time identifier on the continuous time axis and using the time identifier to identify the corresponding position. Optionally, it can be realized by key-value pair, text identifier or pointer, etc. The present application does not limit this. The time point is the position identified by the time identifier on the time axis. The plurality of time points included in the time axis correspond one-to-one to the time identifiers.

[0075] It can be understood that the continuous digital signal set includes a plurality of continuous digital signals, and the plurality of continuous digital signals are continuous and come from the same sensor, so that the continuous digital signal records all values between the start of the signal and the end of the signal. Based on the discretization of the time point, the signal start point and the signal end point of the continuous digital signal are found in the time axis, and all time points between the signal start point and the signal end point are extracted to obtain a target time point set. From the target time point set, the target time point is extracted in turn, and the value corresponding to the target time point is identified in the continuous digital signal, and the value corresponding to the time point is taken as a continuous-discrete signal (i.e. a discrete signal obtained by discretizing a continuous signal). Thus, the continuous digital signal can correspond to a discrete signal group, and the continuous-discrete signal in the discrete signal group corresponds one-to-one to the time point.

[0076] It can be understood that the time stamping of the discrete signal group set is similar to the implementation process of identifying the discontinuous numerical value group set and performing time stamping on each discontinuous numerical value in the discontinuous numerical value group set, and can achieve the same effect. Here, it is not repeated.

[0077] S4, upload the time calibration signal set to the pre-constructed edge processing device to obtain a temporary storage data set, identify the field data set in the temporary storage data set, transmit the field data set to the pre-constructed field device to obtain the field data set, and upload the temporary storage data set to the pre-constructed centralized processing center to obtain the to-be-processed data set.

[0078] It should be noted that the edge processing device is a device for edge computing, which can temporarily store data to increase the stability of data transmission. The field data set is a set of data that needs to be transmitted to the mining site. The field device is a device in the mining site. The centralized processing center is a data control center.

[0079] S5, deleting the temporary data set based on the edge processing device to obtain a cleaned edge processing device.

[0080] Specifically, the deleting the temporary data set based on the edge processing device to obtain a cleaned edge processing device comprises: judging whether the edge processing device completes a preset transmission task, wherein the transmission task is an operation of transmitting the field data set to a pre-constructed field device to obtain a field data set, and uploading the temporary data set to a pre-constructed centralized processing center to obtain a to-be-processed data set; if it is confirmed that the edge processing device completes the transmission task, performing data integrity verification on the field data set and the to-be-processed data set to obtain an integrity verification result; if the integrity verification result is a preset correct transmission, monitoring a total storage time length of the temporary data set in the edge processing device, and if the total storage time length reaches a preset temporary storage time length, deleting the temporary data set in the edge processing device to obtain a cleaned edge processing device.

[0081] It can be understood that the temporary storage time length is a total time length between the edge processing device receiving the temporary data set and deleting the edge processing device, and the temporary storage time length can be set by a person, and the application sets it to 24 hours.

[0082] It can be understood that the performing data integrity verification comprises but is not limited to data consistency checking, timing and continuity checking, etc. The data consistency checking can check whether the temporary data set conforms to a format predefined by the centralized processing center and whether the field data set conforms to a format predefined by the field device. The timing and continuity checking is used to ensure that the field data set and the to-be-processed data set are checked without breakpoints or out-of-order. If the integrity verification result is the preset correct transmission, it indicates that no error occurs in the data transmission process. The data integrity verification can be realized by the prior art, and will not be described here. The total storage time length is a total time length of storing the temporary data set in the edge processing device.

[0083] Specifically, the deleting the temporary data set based on the edge processing device to obtain a cleaned edge processing device further comprises: if it is confirmed that the edge processing device does not complete the transmission task, returning the operation of identifying the field data set in the temporary data set, and performing an increment operation on a preset transmission counter to obtain a transmission number, wherein an initial value of the transmission counter is 0, and the transmission counter is incremented by one each time the operation of returning the field data set in the temporary data set is performed; until it is confirmed that the edge processing device completes the transmission task, the transmission number is greater than or equal to a preset repetition threshold, or the integrity verification result is a preset incorrect transmission; When the integrity check result is a preset incorrect transmission, the operation of returning the field data set in the temporary data set is returned, and an operation of adding one to the pre-constructed transmission count is performed; When the transmission count is greater than or equal to a preset repetition threshold, a communication error warning instruction is generated.

[0084] It can be understood that if the edge processing device does not complete the transmission task, it means that the data transmission is incorrect, and transmission needs to be performed again, so the operation of returning the field data set in the temporary data set is returned. However, when the data transmission is incorrect, it cannot be transmitted endlessly, so the present application performs an operation of adding one to the pre-constructed transmission count to obtain the transmission count, wherein the transmission count is used to calculate the number of times the edge processing device performs the transmission task, and the initial value of the transmission count is 0, and after the edge processing device performs a transmission task, the operation of returning the field data set in the temporary data set is performed once, and an operation of adding one to the transmission count is performed.

[0085] It can be understood that in order to avoid endless transmission, the present application performs an operation of adding one to the transmission count to obtain the transmission count, regardless of whether the integrity check result is a preset incorrect transmission or the edge processing device does not complete the transmission task. The repetition threshold is artificially preset to describe the maximum number of times the edge processing device performs the transmission task. When the transmission count is greater than or equal to the preset repetition threshold, the present application generates a communication error warning instruction to notify the monitoring personnel that a data transmission error has occurred.

[0086] Optionally, the communication error warning instruction is a text instruction that can include error conditions, including but not limited to: the integrity check result is a preset incorrect transmission, or the edge processing device does not complete the transmission task, the transmission count is greater than or equal to the preset repetition threshold, and the like. The specific text can be set by a professional.

[0087] S6, spatiotemporal fusion is performed on the to-be-processed data set to obtain an initial fusion data set.

[0088] It can be understood that the to-be-processed data set is discrete data, and a corresponding time point can be found on the identification time axis, so the operation of performing spatiotemporal fusion is to represent and sort the data at the same time point on the same identification time axis, so that the representation of different sensors in a mining area at the same time point and the time development law can be shown. Therefore, the present application refers to the collection of all data on the identification time axis as an initial fusion data set.

[0089] S7, data cleaning is performed on the initial fusion data set by using the pre-constructed abnormal data identification model, an optimized fusion data set is obtained, and the pre-constructed visualization model set is updated according to the optimized fusion data set.

[0090] It can be understood that the visualization model set is a set composed of a plurality of visualization models, the visualization model can intuitively perceive the spatial and temporal information of the modal mining subsidence data compared with numerical values, and the visualization model can be a three-dimensional model, a two-dimensional numerical fluctuation model, etc., and the application does not limit this. The initial fusion data set can represent the representation of different sensors at the same time and point and the development law over time, but in the application, different sensors are opened at different time periods, so in the processing process, abnormal data may exist in the initial fusion data due to transmission errors of different sensors and opening and closing fluctuations of the sensors. Therefore, the application also needs to perform data cleaning on the initial fusion data set.

[0091] Further, the data cleaning is performed on the initial fusion data set by using the pre-constructed abnormal data identification model, and the optimized fusion data set is obtained, and the pre-constructed visualization model set is updated according to the optimized fusion data set. The initial fusion data is sequentially extracted from the initial fusion data set, and the following operations are performed on the extracted initial fusion data. Data abnormal value identification is performed on the initial fusion data to obtain an abnormal data group, wherein the abnormal data group includes zero, one or more abnormal data. If the abnormal data group is not empty, the abnormal value is removed from the initial fusion data based on the abnormal data group to obtain intermediate fusion data containing one or more missing values. The intermediate fusion data and the pre-constructed mask matrix are input into the pre-constructed missing value generator as input parameters, the output parameters corresponding to the input parameters are predicted by using the missing value generator, and the suboptimal fusion data containing one or more predicted interpolation values are obtained. The suboptimal fusion data is input into the pre-constructed discriminator, the interpolation correct probability of each predicted interpolation value in the suboptimal fusion data is estimated by using the discriminator, one or more interpolation correct probabilities are obtained, the predicted interpolation values corresponding to all interpolation correct probabilities greater than the preset probability threshold in the one or more interpolation correct probabilities are used as available interpolation values, the intermediate fusion data is interpolated, and the optimized intermediate fusion data is obtained. The mask matrix is adjusted based on the optimized intermediate fusion data to obtain an updated adjustment mask matrix, the optimized intermediate fusion data is used as the intermediate fusion data, the updated adjustment mask matrix is used as the mask matrix, and the step of inputting the intermediate fusion data and the pre-constructed mask matrix into the pre-constructed missing value generator as input parameters is returned until the optimized intermediate fusion data does not contain missing values, and the optimized fusion data is obtained. The optimization fusion data is aggregated to obtain an optimization fusion data set.

[0092] It should be noted that the data abnormal value identification is an operation of identifying abnormal values in the initial fusion data, and the abnormal values include but are not limited to outliers deviating from a distribution mode of the initial fusion data, etc. The data abnormal value identification is a prior art, and will not be described here. It should be noted that one initial fusion data includes a plurality of unit fusion data.

[0093] It should be noted that the abnormal data is the abnormal value in the initial fusion data identified by the data abnormal value identification. The abnormal value elimination from the initial fusion data based on the abnormal data set means that all abnormal data in the abnormal data set are eliminated from the initial fusion data. The mask matrix is a matrix used to represent the position of one or more missing values in the intermediate fusion data, and the mask matrix has the same structure as the intermediate fusion data.

[0094] It can be understood that the missing value generator is a neural network constructed based on a convolutional neural network and used for filling the missing values. The process of obtaining the missing value generator includes: The multi-head attention residual block is constructed, wherein the multi-head attention residual block sequentially includes an input layer, a first unit residual block, a second unit residual block, an output layer and a convolution layer, the first unit residual block and the second unit residual block are the same, and the first unit residual block includes an inflation causal convolution layer and a multi-head attention mechanism layer. Based on the pre-constructed residual connection mode, the pre-constructed spatio-temporal convolution network input layer, the plurality of multi-head attention residual blocks and the pre-constructed spatio-temporal convolution network output layer are sequentially connected to obtain an initial temporal convolution network. The generator is constructed according to the pre-constructed initial temporal convolution network, wherein the generator includes a first one-dimensional convolution layer, a second one-dimensional convolution layer, the initial temporal convolution network, a first full connection layer and a second full connection layer.

[0095] It can be understood that the missing value generator in the present application is a network composed of a temporal convolution network, so the input layer, the output layer and the convolution layer are all common structures in the temporal convolution network, which will not be described here. The inflation causal convolution layer is a network structure for processing time series data in the temporal convolution network, which can exponentially increase the receptive field without additional large amount of calculation and parameter amount. The multi-head attention mechanism layer is a network structure introducing a multi-head attention mechanism, which can simultaneously pay attention to the information of different positions of the missing values represented by the mask matrix. It can be understood that the first unit residual block further includes a batch normalization layer, a regularization layer and other conventional prior art for constructing the network structure, which will not be described here.

[0096] Further, the residual connection manner refers to a connection manner of connecting layers based on residual connection. The spatio-temporal convolution network input layer and the spatio-temporal convolution network output layer are input layer and output layer of the time convolution network. Therefore, the initial time convolution network in the application is composed of the spatio-temporal convolution network input layer, the plurality of multi-head attention residual blocks and the spatio-temporal convolution network output layer, and the connection manner between layers is the residual connection manner.

[0097] Specifically, the first one-dimensional convolution layer and the second one-dimensional convolution layer have the same structure, and the first full connection layer and the second full connection layer have the same structure. The first one-dimensional convolution layer is a convolution network structure for processing one-dimensional data, and the first full connection layer is a full connection layer. This technology is prior art and will not be described here. Therefore, in the application, the generator is a network structure sequentially connected by the first one-dimensional convolution layer, the second one-dimensional convolution layer, the initial time convolution network, the first full connection layer and the second full connection layer.

[0098] Further, in the process of predicting the output parameter corresponding to the input parameter by using the missing value generator, the specific structure of the missing value generator has been described in detail in the foregoing. After the input parameter is introduced into the missing value generator, the output parameter is generated, thereby obtaining the suboptimal fusion data containing one or more predicted interpolation values. That is, the generator interpolates the missing part in the intermediate fusion data, the interpolated value is the predicted interpolation value, and the interpolated intermediate fusion data is the suboptimal fusion data.

[0099] It can be understood that the discriminator is a network structure for identifying the probability that one or more interpolation values in the suboptimal fusion data conform to the distribution rule of the suboptimal fusion data. The specific structure can be realized by prior art, and will not be exemplified here.

[0100] Further, the interpolation correct probability is the probability that the interpolation value conforms to the distribution rule of the suboptimal fusion data. The probability threshold is a probability value artificially preset. The predicted interpolation value is screened, and when the interpolation correct probability corresponding to the predicted interpolation value is greater than the preset probability threshold, the application confirms that the predicted interpolation value is a usable interpolation value, and uses the predicted interpolation value to interpolate the intermediate fusion data. Therefore, the usable interpolation value is a value that can be used to interpolate the intermediate fusion data.

[0101] Specifically, the adjusting the mask matrix based on the optimized intermediate fusion data to obtain an updated mask matrix refers to that after obtaining the optimized intermediate fusion data, one or more missing values in the original fusion data may be completely filled or a part of them is not filled. Therefore, a new mask matrix is constructed by using the position containing one or more intermediate missing values in the optimized intermediate fusion data, and the mask matrix is adjusted. Until the optimized intermediate fusion data does not contain missing values, it is indicated that the filling of the fusion data has been completed.

[0102] S8, based on the updated visualization model set, the edge cleaning processing device and the field data set, complete the spatio-temporal fusion of multi-modal mining subsidence data.

[0103] It can be understood that the edge cleaning processing device can provide sufficient space for the next received data. The pre-constructed visualization model set refers to the visualization model set in the historical time, and the updated visualization model set refers to the visualization model set obtained by importing the mining subsidence information collected in the current period to the pre-constructed visualization model set.

[0104] To solve the problems described in the background art, the present application uses a sensor cluster to collect data of different modalities, jointly describes the state of the subsidence of the mining area, and constructs a sensor network to more comprehensively monitor the mining area corresponding to the regional space, and uses a single sensor network that dynamically adjusts the collection precision under the condition of meeting the precision requirement, and closes some sensors under the condition of lower precision requirement, thereby saving energy consumption. The digital signal set is converted into a digital signal set through digital-to-analog conversion, and the digital signal set is time-stamped to obtain a time-stamped signal set. It can be seen that the present application processes digital signals, because digital signals are helpful to realize anti-interference, high precision and programmable processing through binary coding, and provide a standardized data basis for subsequent storage, transmission and intelligent analysis. The digital signal set is time-stamped to obtain a time-stamped signal set. The present application represents different sensors in the same time point in a mining area and the development law over time through a time axis. The time-stamped signal set is uploaded to a pre-constructed edge processing device to obtain a temporary data set, the on-site data set in the temporary data set is identified, the on-site data set is transmitted to a pre-constructed on-site device to obtain an on-site data set, and the temporary data set is uploaded to a pre-constructed centralized processing center to obtain a to-be-processed data set. It can be seen that the present application can temporarily store data to increase the stability of data transmission. The edge processing device deletes the temporary data set based on the edge processing device to obtain a cleaned edge processing device. The present application also performs data integrity verification to check whether the data conforms to the pre-defined format, and ensures that the time series data has no breakpoints or out-of-order. The present application performs an add-one operation on the transmission count to obtain the transmission frequency, whether the integrity verification result is a preset error transmission or the edge processing device does not complete the transmission task. When the transmission frequency is greater than or equal to a preset repetition operation threshold, the present application generates a communication error warning instruction to notify the monitoring personnel that a data transmission error has occurred. Because different sensors are turned on at different times, there may be abnormal data in the initial fusion data due to transmission errors of different sensors and on-off fluctuations of the sensors during processing. Therefore, the present application also performs data cleaning on the initial fusion data set. The present application realizes more accurate monitoring of mining subsidence by fusing the space-time features of multi-dimensional data.

[0105] As Figure 2 shown is a functional module diagram of a multi-modal mining subsidence data space-time fusion system provided by an embodiment of the present application.

[0106] The multi-modal mining subsidence data space-time fusion system 100 can be installed in an electronic device. According to the functions implemented, the multi-modal mining subsidence data space-time fusion system 100 can include a pre-processing data module 101, a time calibration module 102, an uploading module 103, and a post-processing module 104. The modules of the present application can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete a fixed function, which are stored in the memory of the electronic device.

[0107] The pre-processing data module 101 is configured to construct a sensor network and extract a sensor cluster from the sensor network, wherein the sensor cluster includes a plurality of sensors, collect mine subsidence information using the sensor cluster, and pre-process the mine subsidence information to obtain a digital signal set, wherein the digital signal set includes a discrete signal set and a continuous signal set. The time calibration module 102 is configured to perform timestamp calibration on the digital signal set to obtain a time-calibrated signal set. The uploading module 103 is configured to upload the time-calibrated signal set to a pre-constructed edge processing device to obtain a temporary data set, identify an on-site data set in the temporary data set, transmit the on-site data set to a pre-constructed on-site device to obtain an on-site data set, and upload the temporary data set to a pre-constructed centralized processing center to obtain a to-be-processed data set. The edge processing device is deleted based on the edge processing device to obtain a cleaned edge processing device. The post-processing module 104 is configured to perform space-time fusion on the to-be-processed data set to obtain an initial fusion data set, use a pre-constructed abnormal data identification model to perform data cleaning on the initial fusion data set to obtain an optimized fusion data set, update a pre-constructed visualization model set according to the optimized fusion data set, and complete the multi-modal mining subsidence data space-time fusion based on the updated visualization model set, the cleaned edge processing device, and the on-site data set.

[0108] In detail, the modules in the multi-modal mining subsidence data space-time fusion system 100 in the embodiments of the present application use the same technical means as the multi-modal mining subsidence data space-time fusion method described in the above Figure 1 , and can produce the same technical effects, which will not be described here.

[0109] As Figure 3 shown is a structural schematic diagram of an electronic device for implementing the multi-modal mining subsidence data space-time fusion method according to an embodiment of the present application.

[0110] The electronic device 1 can include a processor 10, a memory 11 and a bus 12, and can further include a computer program stored in the memory 11 and executable on the processor 10, such as a multi-modal mining subsidence data space-time fusion method program.

[0111] The memory 11 includes at least one type of readable storage medium, such as a flash memory, a mobile hard disk, a multimedia card, a card-type memory (e.g., an SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 11 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the memory 11 can include both an internal storage unit and an external storage device of the electronic device 1. The memory 11 can be used not only to store application software and various data installed in the electronic device 1, such as a multi-modal mining subsidence data space-time fusion method program code, but also to temporarily store data that has been output or will be output.

[0112] The processor 10 can be composed of an integrated circuit in some embodiments, such as a single packaged integrated circuit or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, combinations of various control chips, etc. The processor 10 is the control core of the electronic device, which connects various components of the electronic device through various interfaces and lines, executes or runs programs or modules stored in the memory 11 (such as a multi-modal mining subsidence data space-time fusion method program, etc.), and calls data stored in the memory 11 to perform various functions and process data of the electronic device 1.

[0113] The bus 12 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus 12 can be divided into an address bus, a data bus, a control bus, etc. The bus 12 is configured to enable connection and communication between the memory 11, the at least one processor 10, etc.

[0114] Figure 3 Only the electronic device with components is shown, and those skilled in the art can understand that, Figure 3 The structure shown does not constitute a limitation on the electronic device 1, and can include fewer or more components than shown, or combine certain components, or different component arrangements.

[0115] For example, although not shown, the electronic device 1 can also include a power supply (such as a battery) to power each component. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management system, so that the power management system can implement functions such as charge management, discharge management, and power consumption management. The power supply can also include one or more direct current or alternating current power sources, a recharging system, a power failure detection circuit, a power converter or inverter, a power status indicator, etc. The electronic device 1 can also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which are not described here.

[0116] Further, the electronic device 1 can also include a network interface, which can optionally include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), and is typically used to establish a communication connection between the electronic device 1 and other electronic devices.

[0117] Optionally, the electronic device 1 can also include a user interface, which can be a display (Display), an input unit (such as a keyboard (Keyboard)), and optionally a standard wired interface, a wireless interface. Optionally, in some embodiments, the display can be an LED display, a liquid crystal display, a touch liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, etc. The display can also be appropriately referred to as a display screen or a display unit, and is used to display information processed in the electronic device 1 and to display a visualized user interface.

[0118] The multi-modal mining subsidence data space-time fusion method program stored in the memory 11 in the electronic equipment 1 is a combination of multiple instructions, which can realize the following when running in the processor 10: constructing a sensor network and extracting a sensor cluster from the sensor network, wherein the sensor cluster comprises a plurality of sensors; collecting mine subsidence information by using the sensor cluster and preprocessing the mine subsidence information to obtain a digital signal set, wherein the digital signal set comprises a discrete signal set and a continuous signal set; performing timestamp calibration on the digital signal set to obtain a time-calibrated signal set; uploading the time-calibrated signal set to a pre-constructed edge processing device to obtain a temporary data set, identifying an on-site data set in the temporary data set, transmitting the on-site data set to a pre-constructed on-site device to obtain an on-site data set, and uploading the temporary data set to a pre-constructed centralized processing center to obtain a to-be-processed data set; deleting the temporary data set based on the edge processing device to obtain a cleaned edge processing device; performing space-time fusion on the to-be-processed data set to obtain an initial fusion data set; performing data cleaning on the initial fusion data set by using a pre-constructed abnormal data identification model to obtain an optimized fusion data set, and updating a pre-constructed visualization model set according to the optimized fusion data set; based on the updated visualization model set, the cleaned edge processing device and the on-site data set, completing the multi-modal mining subsidence data space-time fusion.

[0119] Specifically, the specific implementation method of the processor 10 to the above instructions can refer to Figures 1 to 3 The description of related steps in the corresponding embodiments is not repeated here.

[0120] Further, the modules / units integrated in the electronic equipment 1 are implemented in the form of software function units and sold or used as independent products, which can be stored in a computer readable storage medium. The computer readable storage medium can be volatile or non-volatile. For example, the computer readable medium can include any entity or system capable of carrying the computer program code, a recording medium, a U disk, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory).

[0121] The application also provides a computer readable storage medium, which stores a computer program, and the computer program can realize the following when executed by a processor of an electronic equipment: Construct a sensor network, and extract a sensor cluster from the sensor network, wherein the sensor cluster comprises a plurality of sensors; Collect mine subsidence information by using the sensor cluster, and preprocess the mine subsidence information to obtain a digital signal set, wherein the digital signal set comprises a discrete signal set and a continuous signal set; Perform timestamp calibration on the digital signal set to obtain a time-calibrated signal set; Upload the time-calibrated signal set to a pre-constructed edge processing device to obtain a temporary storage data set, identify an on-site data set in the temporary storage data set, transmit the on-site data set to a pre-constructed on-site device to obtain the on-site data set, and upload the temporary storage data set to a pre-constructed centralized processing center to obtain a to-be-processed data set; Delete the temporary storage data set based on the edge processing device to obtain a cleaned edge processing device; Perform spatio-temporal fusion on the to-be-processed data set to obtain an initial fusion data set; Perform data cleaning on the initial fusion data set by using a pre-constructed abnormal data identification model to obtain an optimized fusion data set, and update a pre-constructed visualization model set according to the optimized fusion data set; Complete spatio-temporal fusion based on multi-modal mining subsidence data based on the updated visualization model set, the cleaned edge processing device and the on-site data set.

[0122] In several embodiments provided in the present application, it should be understood that the disclosed devices, systems and methods can be implemented in other ways. For example, the above-described system embodiments are only illustrative, and actual implementation can have another division way.

[0123] The modules described as separate components can or can not be physically separated, and the components shown as modules can or can not be physical units, i.e., they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs.

[0124] In addition, the functional modules in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software functional modules.

[0125] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0126] Finally, it should be noted that the above examples are merely intended to illustrate the technical solutions of the present application and not to limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A spatiotemporal fusion method based on multimodal mining subsidence data, characterized in that, The method includes: Construct a sensor network and extract sensor clusters from the sensor network, wherein the sensor clusters include multiple sensors; A sensor cluster is used to collect mine subsidence information, and the mine subsidence information is preprocessed to obtain a digital signal set, which includes a discrete signal set and a continuous signal set. Perform timestamp calibration on the digital signal set to obtain a timestamp signal set; The time-calibrated signal set is uploaded to a pre-built edge processing device to obtain a temporary dataset. The field dataset in the temporary dataset is identified, and the field dataset is transmitted to a pre-built field device to obtain a field dataset. The temporary dataset is then uploaded to a pre-built centralized processing center to obtain a dataset to be processed. The temporary dataset is deleted based on the edge processing device to obtain the cleaned edge processing device; Spatiotemporal fusion is performed on the dataset to be processed to obtain an initial fused dataset; Using a pre-built anomaly data identification model, the initial fused dataset is cleaned to obtain an optimized fused dataset, and the pre-built visualization model set is updated based on the optimized fused dataset. Spatiotemporal fusion of multimodal mining subsidence data was completed based on the updated visualization model set, edge clearing processing device, and field dataset.

2. The spatiotemporal fusion method based on multimodal mining subsidence data as described in claim 1, characterized in that, The construction of the sensor network includes: Obtain the target monitoring parameter set and confirm the regional spatial model corresponding to the target monitoring parameter set. Extract the target monitoring parameters sequentially from the target monitoring parameter set, and perform the following operations on each extracted target monitoring parameter: One or more key layout points are identified from the area spatial model. Based on preset deployment distance control parameters, multiple placeable points are identified from the area spatial model. Key layout points are extracted sequentially from the one or more key layout points, and the following operations are performed on the key layout points: The key layout points are combined with multiple placeable points to obtain multiple pairing groups. The spatial Euclidean distance of each pairing group in the multiple pairing groups is calculated to obtain the distance set. Summarize the distance set, and extract all distances greater than the control parameter of the deployment distance from the summarized distance set to obtain multiple suboptimal distances. Identify the suboptimal deployment point corresponding to each suboptimal distance among the multiple suboptimal distances to obtain multiple suboptimal deployment points. Based on the preset sensor layout density gradient, multiple precision layout groups are constructed using one or more key layout points and multiple suboptimal control points. According to the preset time period precision requirements, the multiple precision layout groups are matched for time period precision requirements to obtain multiple precision requirement matching groups. Based on multiple accuracy requirement matching groups, a single-sensor network with dynamically adjusted acquisition accuracy based on time period is constructed. By summarizing the single-sensor networks, the sensor network corresponding to the target monitoring parameter set is obtained.

3. The spatiotemporal fusion method based on multimodal mining subsidence data as described in claim 2, characterized in that, The method, based on a preset sensor layout density gradient, constructs multiple precision layout groups using one or more key layout points and multiple suboptimal control points, including: The target accuracy gradient is extracted from the sensor layout density gradient, wherein the sensor layout density gradient includes: high density gradient, medium density gradient and low density gradient; Based on the target accuracy gradient, multiple target sensors for monitoring target monitoring parameters are identified. Multiple key sensors for setting at the one or more key layout points are extracted from the multiple target sensors to obtain a key sensor group. The key sensor group includes one or more key sensors, and the key sensors correspond one-to-one with the key layout points. From multiple target sensors, key sensor groups are removed to obtain multiple remaining sensors. Based on the multiple remaining sensors and the multiple suboptimal control points, the total number of traversal strategies is calculated. Based on the total number of traversal strategies, multiple traversal layout strategies are confirmed. The total number of traversal strategies is equal to the total number of multiple traversal layout strategies. The formula for calculating the total number of traversal strategies is as follows: ， in, This indicates the total number of traversal strategies. This represents the total number of suboptimal control points. This represents the total number of remaining sensors. Represents the factorial symbol; The traversal layout strategy is extracted sequentially from multiple traversal layout strategies. The detection range coverage, average sensor accuracy, and average network latency of the traversal layout strategy are calculated to obtain the strategy evaluation parameters. Based on the pre-built evaluation method and strategy evaluation parameters, the strategy evaluation value of the traversal layout strategy is calculated. Summarize the strategy evaluation values, and extract one or more strategy evaluation values ​​with the largest values ​​from the summarized strategy evaluation values ​​to obtain one or more suboptimal evaluation values. If there are multiple suboptimal evaluation values ​​with the same value, calculate multiple dispersions corresponding to the multiple suboptimal evaluation values ​​with the same value, and confirm the traversal layout strategy corresponding to the suboptimal evaluation value with the largest dispersion among the multiple dispersions as the suboptimal control point layout strategy of the target accuracy gradient. If there is a suboptimal evaluation value, then the traversal layout strategy corresponding to the suboptimal evaluation value is confirmed as the suboptimal control point layout strategy of the target accuracy gradient. A precision layout group is generated based on a suboptimal control point layout strategy and one or more key layout points to produce a target precision gradient. Summarize the precision layout groups to obtain multiple precision layout groups.

4. The spatiotemporal fusion method based on multimodal mining subsidence data as described in claim 3, characterized in that, The method involves using a sensor cluster to collect mine subsidence information and preprocessing the information to obtain a digital signal set, including: Extract the target sensors sequentially from the sensor cluster, and perform the following operations on each extracted target sensor: Based on the preset sensor sampling frequency of the target sensor, the pre-constructed target sampling information is acquired to obtain the acquired signal. Then, the acquired signal is denoised using a pre-constructed wavelet transform filtering operation to obtain the denoised signal. The wavelet transform filtering operation includes multiple correction filtering calibration parameters, which are obtained by using a pre-constructed orthogonal experimental method to perform parameter correction operations based on the target sensor. The multiple correction filtering calibration parameters are: wavelet basis function, wavelet threshold, and number of decomposition layers. The denoised signals are aggregated to obtain the denoised signal set corresponding to the sensor cluster. The denoised signal set is then subjected to signal classification based on signal continuity discrimination to obtain the digital signal set.

5. The spatiotemporal fusion method based on multimodal mining subsidence data as described in claim 4, characterized in that, The process of performing timestamp calibration on the digital signal set to obtain a timestamp signal set includes: Discrete signal sets are identified from digital signal sets, and the discrete signal sets are divided based on sensor clusters to obtain discontinuous value sets. Each discontinuous value in the discontinuous value set is timestamped to obtain an identified discontinuous value set. The timestamping includes: sensor identifier, monitoring point identifier, and time identifier. Obtain the continuous digital signal set corresponding to the continuous signal set, and perform timestamp calibration on the continuous digital signal set based on the identified discontinuous numerical set to obtain the identified continuous numerical set. By summing up the sets of continuous and discontinuous numerical values, a time calibration signal set is obtained.

6. The spatiotemporal fusion method based on multimodal mining subsidence data as described in claim 5, characterized in that, The process of timestamping a continuous digital signal set based on an identifier discontinuous numerical set to obtain an identifier continuous numerical set includes: Extract a set of time identifiers from the set of discontinuous numerical data, identify the earliest start time and the latest end time in the set of time identifiers, construct a continuous time axis based on the start time and end time, and use the time identifiers in the set of time identifiers to anchor points on the continuous time axis to obtain the identifier time axis, wherein the identifier time axis includes multiple time points. Based on the multiple time points, each continuous digital signal in the continuous digital signal set is discretized based on the time points to obtain a discrete signal set, wherein the discrete signal set corresponds one-to-one with the continuous digital signal, and the discrete signal set includes multiple continuous-discrete signals. The discrete signal set is timestamped to obtain a set of continuously identified numerical values.

7. The spatiotemporal fusion method based on multimodal mining subsidence data as described in claim 6, characterized in that, The step of deleting the temporary dataset based on the edge processing device to obtain the cleaned edge processing device includes: Determine whether the edge processing device has completed the preset transmission task, wherein the transmission task is the operation of transmitting the field dataset to the pre-built field device to obtain the field dataset, and uploading the temporary dataset to the pre-built centralized processing center to obtain the dataset to be processed; If it is confirmed that the edge processing device has completed the transmission task, then the data integrity of the on-site dataset and the dataset to be processed is checked to obtain the integrity check result. If the integrity check result is a preset correct transmission, then monitor the total storage time of the temporary dataset in the edge processing device. If the total storage time reaches the preset temporary storage time, then delete the temporary dataset in the edge processing device to clean up the edge processing device.

8. The spatiotemporal fusion method based on multimodal mining subsidence data as described in claim 7, characterized in that, The process of deleting the temporary dataset based on the edge processing device to obtain a cleaned edge processing device further includes: If it is confirmed that the edge processing device has not completed the transmission task, the operation of returning the field dataset in the identification temporary dataset is performed, and the pre-built transmission count is incremented by one to obtain the number of transmissions. The initial value of the transmission count is 0, and the transmission count is incremented by one every time the operation of returning the field dataset in the identification temporary dataset is performed. The transmission continues until the edge processing device completes the transmission task, the number of transmissions is greater than or equal to a preset repetitive operation threshold, or the integrity check result is a preset erroneous transmission. If the integrity check result is a preset error transmission, then return to the operation of returning the on-site dataset in the temporary identification dataset, and perform an increment operation on the pre-built transmission count; If the number of transmissions is greater than or equal to a preset repetitive operation threshold, a communication error warning instruction is generated.

9. The spatiotemporal fusion method based on multimodal mining subsidence data as described in claim 8, characterized in that, The process of using a pre-built anomaly data identification model to clean the initial fused dataset to obtain an optimized fused dataset includes: Extract initial fusion data sequentially from the initial fusion dataset, and perform the following operations on each extracted initial fusion data: The initial fused data is subjected to outlier identification to obtain anomaly data groups, wherein the anomaly data groups include zero or one or more anomaly data. If the abnormal data group is not an empty set, then the initial fused data is subjected to outlier removal based on the abnormal data group to obtain intermediate fused data containing one or more missing values. The intermediate fused data and the pre-constructed mask matrix are used as input parameters and imported into the pre-constructed missing value generator. The missing value generator is used to predict the output parameters corresponding to the input parameters to obtain suboptimal fused data containing one or more predicted imputation values. The suboptimal fused data is imported into a pre-built discriminator. The discriminator is used to estimate the interpolation correctness probability of each predicted interpolation value in the suboptimal fused data, and one or more interpolation correctness probabilities are obtained. The predicted interpolation values ​​corresponding to all interpolation correctness probabilities greater than a preset probability threshold among the one or more interpolation correctness probabilities are used as available interpolation values ​​to interpolate the intermediate fused data, thereby obtaining optimized intermediate fused data. Based on the optimized intermediate fused data, the mask matrix is ​​adjusted to obtain the updated adjusted mask matrix. The optimized intermediate fused data is used as the intermediate fused data, and the updated adjusted mask matrix is ​​used as the mask matrix. The step of importing the intermediate fused data and the pre-constructed mask matrix as input parameters into the pre-constructed missing value generator is returned until the optimized intermediate fused data does not contain missing values, and the optimized fused data is obtained. The optimized and merged data is aggregated to obtain the optimized and merged dataset.

10. A spatiotemporal fusion system based on multimodal mining subsidence data, characterized in that, The system includes: A preprocessing data module is used to construct a sensor network and extract a sensor cluster from the sensor network. The sensor cluster includes multiple sensors. The sensor cluster is used to collect mine subsidence information and preprocess the mine subsidence information to obtain a digital signal set, which includes a discrete signal set and a continuous signal set. The time calibration module is used to perform timestamp calibration on the digital signal set to obtain a time-calibrated signal set; The upload module is used to upload the time-calibrated signal set to the pre-built edge processing device to obtain a temporary dataset, identify the field dataset in the temporary dataset, transmit the field dataset to the pre-built field device to obtain a field dataset, upload the temporary dataset to the pre-built centralized processing center to obtain a dataset to be processed, and delete the temporary dataset based on the edge processing device to obtain a cleaned edge processing device. The post-processing module is used to perform spatiotemporal fusion of the dataset to be processed to obtain an initial fused dataset. It then uses a pre-built anomaly data identification model to clean the initial fused dataset to obtain an optimized fused dataset. Based on the optimized fused dataset, it updates the pre-built visualization model set. Finally, based on the updated visualization model set, the edge cleaning device, and the on-site dataset, it completes the spatiotemporal fusion of multimodal mining subsidence data.

Images